AI notice
Parlozo's voice teacher is an artificial intelligence. This page says exactly what that means — the notices you get, what the AI never does, and how you control it.
1. When you are told
You are told that you are talking to an AI at three moments: at sign-up, in the first message of every voice conversation (the notice is SPOKEN, in the conversation's language) and every 3 hours in long sessions. This is required by the European AI Regulation (art. 50) and similar laws in other countries where we operate.
2. What the AI never does
- It does not pass as a person: it introduces itself as AI.
- It does not imitate a real person's voice and does not clone YOUR voice.
- It does not identify who is speaking (no voiceprint).
- It is not the sole judge of your progress: lessons are graded by a deterministic, auditable engine that is not the AI.
- It does not store sensitive data about you (health, religion, sexual orientation, finances, documents) — server-side filter.
3. AI-generated content
Shareable cards generated in the conversation use fixed, pre-approved templates with no profanity — and carry a visible “AI” label plus provenance metadata (C2PA/IPTC) marking the content as AI-generated.
4. Moderation and hard moments
The model's output is moderated against sexual content, hate and illegality. If a risk signal to your health appears (self-harm or suicidal ideation), the teacher changes stance immediately and shows the help line of the country you are in (in Brazil, CVV 188). We only record the aggregate count of such events, never content.
5. AI controls that are yours
- Teacher memory: own opt-in, sensitive-data filter, editing and erasure under Profile → Teacher memory.
- Voice processing consent: onboarding and a profile switch.
- Export and deletion of your data: Profile → Your data.